This blog post comes on the heels of some cool project work we’ve been doing with Enterra Solutions® surrounding how its cognitve computing solution can contribute to SAP Integrated Business Planning to ring further business insights from master data as well as other relevant structured and unstrucutured large data sets. I’m grateful to collaborate with Stephen DeAngelis, founder and CEO of Enterra Solutions to compose this post and to dive a bit into understanding how machine learning and cognitive computing relate.
I do a fair amount of research, rewriting, and editing as part of my blogging regiment. The majority of the projects we enable and run in the COIL are underscored by a multitude of technologies; cloud platforms, virtualization, augmented reality, mobility, cyber security, big data analytics, artificial intelligence (AI), and Machine Learning (ML).
These latter two topics has certainly tested my blogging rigor in preparation to describe our latest project with Enterra Solutions® and how to come to understand how cognitive computing can be used to gain deeper insights into a business.
We begin with discussing where we are at today in a straightforward project exploring how the Enterra® Enterprise Cognitive System™ (ECS) and its Supply Chain Intelligence Solution (SCIS) can be leveraged to inform a supply chain planner’s perspective using SAP Integrated Business Planning (IBP) Control Tower with rich business insights relative to supply chain planning and optimization.
With consideration for the project team’s desired and expected outcome forthis project, this blog attempts to characterize how ECS is situated as a complement to SAP IBP and what a complete integration of ECS and SAP HANA might offer. I am amped to welcome Stephen to help fully inform on the known and potential complements and how machine learning inputs to cognitive computing systems are of value, This project has been in flight at the lab just ahead of other recent news that SAP has Introduced Machine Learning and Predictive Analytics to Cloud-Based Integrated Business Planning Suite:
IBP will clearly take full advantage of insights drawn from machine learning embedded into the suite. This embedded Machine Learning is what can make applications smarter and capable of acting on what the algorithm learns or pattern it detects from data in order to deliver a prediction.
What we hope to observe as a part of the current project, is how the two systems can become highly complementary. Our COIL project seeks evidence of being able to accurately inject valid insights into the IBP Control Tower from Enterra ECS. Insights formulated from mathematical analysis, optimization and semantic reasoning performed on supply chain master and external data. Enterra’s ECS cognitive computing acts beyond machine learning by leveraging a Representational Learning Machine™ (RLM) created by Massive Dynamics that includes functional mathematical representations, nonlinear mappings, and high degrees of interdependencies. Insights are generated through semantic and mathematical analysis and then executed through the system of record from automation and optimization.
The View from COIL: A Brief History of Co-Innovation Research
Tech topics evolve rapidly, and from our vantage point from within COIL, we have the good fortune to regulary obtain a hands-on perspective of what cutting/bleeding edge technology drives IT and computer technology trends. Ten years ago, enterprise service-oriented architecture was topic du jour in the COIL Palo Alto lab. Nearly five years ago, SAP HANA was introduced generating serious focus upon the relevance and importance of big data, semantic text analytics, conditioned based maintenance and predictive analytics. SAP Co-innovation project work spans all of these realms.
Less than three years ago, the hyperbole and buzz surrounding the Internet of Things had nearly turned the acronym IoT into a regularly spoken mantra across many lines of business as well as a call to action to learn from all this data. IoT can and does already generate massive amounts of data and rapidly maturing machine learning is fundamental to help analyze it. Although the IoT hasn’t fully emerged, in scores of industries, algorithms have been purposely designed to endlessly feed upon more and more sensor supplied data; much of it streamed real time.
COIL is observing marked interest today in the topic of Artificial Intelligence. We have today both active and developing projects working to leverage machine learning and cognitive computing. This real uptick in AI project work is interesting considering that this is an area of technology that has been in development for decades. Commodity computing and lower costs to collect and store data are two accelerants establishing AI as a new cornerstone to what is the art of the possible. Evidence now abounds for how Machine learning algorithms (the most basic form of AI) are being successfully used to look for patterns that a human could not hope to detect when manually trying to analyze some massive data set.
The Rise of Cognitive Computing
There is a fast-emerging opportunity to leverage this pattern detection and go further to emulate human thought process and to determine the best way to support human decision making — up to and including machines making decisions on behalf of humans. This may sound far in the distance to some, but the AI horizon draws nearer each day. On the heels of this first co-innovation project with Enterra, there is an opportunity to explore how to use machine learning output as input to a cognitive engine. Seth Earley (@sethearley), CEO of Earley Information Science, writes, “Machine learning is a technique for detecting patterns and surfacing information, using many different mechanisms based on statistics and mathematical models. … Machine learning algorithms pull back information that’s relevant to the user by looking for patterns, improving the search results. … Cognitive computing is a newer, emerging field. It’s about making computers more user friendly, with an interface that understands more of what the user wants”.
Unlike machine learning systems, cognitive computing systems can be formed to address many more variables than previous analytic platforms and can be designed to integrate both structured and unstructured data to render complete analysis. There is plenty to learn with respect to solving difficult business problems using AI, but even with obtaining confidence in a given algorithm to provide insight or to solve a problem, there are plenty of other challenges surrounding successful use. Although the term “cognitive computing” seems to imply the system is sentient, as we explain later, that’s not the case. Cognitive computing falls in the category known as narrow Artificial Intelligence.
AI typically requires a lot of computing power. It is advantageous to explore different sets of data from different sources at the same time. As Tim Allen (@TimIntel), Global SAP Marketing Alliance Manager at Intel with Intel, describes it, “Analyzing these massive datasets requires new technologies, flexible cloud computing, and virtualization software such as Apache Hadoop and Spark. It also needed more powerful, high-performance processors that provided the tools to uncover the insights in Big Data”. There is perhaps no better way to examine how to apply these new technologies than in a co-innovation lab environment capable of replicating the scale up and scale out architectures AI solutions can demand.
Digital Transformation and Cognitive Computing: An Industry View
Digital transformation is a goal for most firms today and it’s driving the AI and Cognitive Computing topic. An IDC report from October 2016 forecasts that adoption of cognitive systems and artificial intelligence across a broad range of industries will drive global revenues from nearly $8.0 billion in 2016 to more than $47 billion in 2020. IDC also projects the market for cognitive/AI solutions will experience a compound annual growth rate (CAGR) of 55.1% from 2016-2020. This growth does suggest some humans have a solid sense for what AI means to our future. Even so, when it comes to particulars, you may find not everyone is clear on their understanding of the key concepts and working definitions of artificial intelligence and cognitive computing.
In the COIL Silicon Valley, the project participants have spent a good deal of time into trying to understand these topics well enough to help enable other proposed co-innovation projects. The expectation in the market and with customers is that AI makes machines smart. We’ve arrived at such an expectation quite rapidly given that up until not so long ago, business was driven only by time-series business statistics. What’s cool about our project is that it fits well with Vanguard Software’s John Hayes’ assertion that “supply chain is perhaps the industrial epicenter of analytics innovation, and has historically been a major force in the advancement of all manner of predictive technology, including machine learning”. While we see us in the very beginnings of project work to explore SAP planning and how use of its embedded machine learning with cognitive computing from Enterra might bring penetrating insights and benefits to a supply chain planner, the work is contributing to the corpus of analytics, ML and AI that helps advance firms towards an entirely digital and autonomous supply chain.
As we’ve come to understand the perspective of the planner it would seem one fundamental goal is to automate solving problems without requiring human assistance. A planner solving multiple tiers of simple or complex problems is time not contributed to planning optimization. We have an opportunity (through ongoing co-innovation project work) to discover how a cognitive computing system like ECS can use machine learning algorithms embedded within the SAP suite. What we know from talking to experienced planners is that generic answers are not useful,
Can ECS find the underlying functional drivers that explain and control the system from process data in order to anticipate new problems and perform optimization with a complex objective function that intelligently weights competing goals? It’s fortunate that we can explore a co-innovated approach to address some very big questions. The reality is that advances in compute methods to support supply chain operations has focused squarely on low-level operations and not upon the rudiments of a planner saddled with product forecasting and strategy. Supply chain planners (i.e., humans) may not trust a machine to make even ad hoc decisions; but, in time, humans will establish such trust. Now is the time to understand what is required for this trust to become forged.
As noted above, machine learning is the most basic of what is called Artificial Intelligence today. AI is all about what it means for a machine to be viewed as intelligent. ML enables AI. Its focus is to automate discovery of regularities in data and then generalize this into new but similar data. Keeping brevity in mind, there are other dimensions of AI and Machine Learning we could weave into our discussion — like deep learning (i.e., neural nets), statistical learning theory, or natural language processing — but our aim here to look more closely at the dimension of AI referred to as cognitive computing. Read enough articles on the term and you will begin to understand cognitive computing focuses upon the ability of a machine to reason and understand at a higher level, like a human. This includes more symbolic and conceptual information rather than just learning from data through using non-brittle inference reasoners and common sense, industry specific, and operational specific ontologies.
Cognitive computing is at the rich far end of the AI spectrum from traditional machine learning algorithms. It has both a flexible framework and architecture composed of computational intelligence (machine learning and other sophisticated mathematical analysis techniques) and semantic reasoning integrated into a functioning platform. Maybe it’s too soon to establish a concrete definition? Some stay away from the term altogether or use a proxy term like computational intelligence. For the sake of this discussion I will stick with Enterra’s definition of cognitive computing which is:
Cognitive Computing combines Semantic Intelligence (i.e., artificial intelligence + natural language processing + ontologies) with Computational Intelligence (i.e., advanced mathematics).
Below are a few terms that explain how cognition plays a role in cognitive computing.
Cognition: The set of internal states, mechanisms that allow a cognitive agent (human, animal or some machines) to manifest reasoning, logic, memory call, etc.
Cognitive computing: Extending cognition to something more than input, process, output (the traditional AI) to target reasoning and logic.
Cognitive computing systems use machine learning algorithms. Such systems continually acquire knowledge from the data fed into them by mining data for information. The systems refine the way they look for patterns and as well as the way they process data so they become capable of anticipating new problems and modeling possible solutions.
Let me now turn over to my co-blogger Stephen to provide further clarity.
The Enterra Enterprise Cognitive System™
The Enterra Enterprise Cognitive System is designed to provide insightful analyses of high volumes of data. The technology uses mathematical and statistical analysis combined with artificial intelligence to create “an efficient and meaningful approach” to analyzing vast amounts of data. Its computational intelligence engines are designed with the acknowledgement that the size of data can be significant and that the data can be complex and highly dimensional (including high order interdependencies, non-parametric, and non-linear relationships). It additionally provides semantic intelligence to represent and understand subtle knowledge and to render an ability to both reason and infer. ECS is comprised of a number of state of the art technologies including:
- Common Sense and Industry Rules-based Ontologies — The ECS uses the Cyc platform for both its world’s largest Commonsense Ontology as well as its non-brittle rule engine.
- Rules-based Inference System — A set of code and mathematical logic that allows for forward chaining and backward chaining of rules and logic to reason about data much like a human would. This allows the ECS to determine what is non-obvious and provides the ability to suggest potentially confounding variables not included within the original analysis but whose absence could result in misleading conclusions.
- Hypothesis Engine. A set of rules and code allowing the system to diagnose a problem, determine a reasoning and computational plan, iterate with advanced math engines, and then interpret results, take actions, and learn from the results.
- Mathematical and Quantitative Analytics Engines — Along with traditional analytic approaches, the ECS uses Massive Dynamics’ Representational Learning Machine, which is usually superior to other computer learning techniques due to its ability to optimally fit high-dimensional, higher-order, non-linear functions to smaller-size data sets with a precise measurement of influence by variable.
- Natural Language Processing — Allows the ECS to communicate with users in a non-technical way.
We can discuss the ECS and its technical capabilities within a broader context of how supply chain experts can apply cognitive computing to help integrate data across an enterprise and ignore the fact that data is sourced from multiple silos. This capability provides all decision makers with access to appropriate data and helps foster corporate alignment.
Enterra at COIL
The COIL project in flight today assumes the persona of a supply chain planner in the team’s effort to show how the ECS and the SAP IBP Control Tower can be used to generate business insights. In this effort, they (i.e., the planner personas) are using master data provided by a major consumer packaged goods customer. The team’s focus is to address data Integration needs and to deep dive on communication and messaging protocols between IBP and ECS. The Master Data Integration is now completed. The major team activity will be integrating Alerts related to the tables in SAP HANA.
This project’s objective is to integrate an instance of Enterra ECS on SAP HANA to an SAP cloud instance of IBP so we can demonstrate how these insights get directed to the planner in the form of alerts and suggestions useful for decision support applied to ad hoc situations occurring across a company’s complex supply chain. Time and again it has been shown the faster a company can react to an alert the more effective the response.
Enterra ECS and SAP
As we close in on the primary goal of this first project, it affords us an opportunity to discover more specifically how cognitive computing can lead to outright autonomous supply chains. Supply chains will change dramatically in this decade; but, it’s rare for complex systems to become profoundly changed overnight. Now is the time to begin ideating and developing more project work to render a desirable co-innovated solution between Enterra ECS, SAP HANA and SAP IBP. Continuous co-innovation project work can be done to capture proof points for how such a co-innovated solution might be implemented and managed to perform sophisticated mathematical analysis, pattern recognition and automated expert planning functions better than humans when it comes to supply chain management.
Cognitive computing — and how it is applied to supply chain management — is a red hot topic and, as such, a way in which to characterize such a system has emerged.
- Interpret and vary responses based on context
- Personalize the results or response based on multiple sensor inputs
- Learning and improving from prior experience
- Pattern Detection within very large data sets
- Prediction of results and actions
- Render logical inference
There are other attributes of interest — like how a cognitive system deals with uncertainty or dirty data or how to produce results from combining multiple data sets. For the sake of this discussion, we can consider the aforementioned with respect to how ECS and SAP HANA together might become an “intelligence-based” system that is probabilistic versus deterministic, contextual as well as becoming a system that learns based upon new information and ongoing interactions with users. Demand planning and Inventory optimization are just two dimensions of supply chain management where the co-innovation work we do today helps to quickly illuminate the path leading to how we find balance between what humans and machines do to obtain supply chain optimization.
“Big data and analytics already allow supply chain managers greater insight into their operations,” writes Abe Eshkenazi (@aeshkenazi), CEO at APICS, “which enables increased efficiencies and greater data-driven insights. The number of data sources in the supply chain is expected to grow exponentially, which will offer managers even more opportunity for contextual intelligence and more knowledge sharing across and between organizations. For example, with additional data, planning and decision-making will improve across an organization, which can lead to decreased costs and time spent on operations.
Data-driven insights have the power to decrease risk, improve transparency, identify trends, and initiate automatic responses where it wasn’t possible before. The supply chain of the future will build on these benefits to become even more streamlined and efficient than it is now. However, it’s imperative that organizations are open to learning new processes for harnessing and interpreting this data, otherwise they may face intense pressure from more agile competitors.”
In the past, it took some imagination to envision a cognitive supply chain because discussions about cognitive computing, digital transformation, , and the Internet of Things (IoT) were still fairly new. As these technologies have matured the picture has become much clearer. Lora Cecere (@lcecere), founder and CEO of Supply Chain Insights writes, “Cognitive learning’s potential to drive value is exciting. … Supply chains will learn as we sleep, and the insights will be deeper and richer with fewer people. It is coming. It is real. Are you ready?”
In a recent report, Gartner analysts note, “Enterra Solutions offers AI capabilities to solve supply chain problems. Through deep learning, rule-based inferences, natural-language processes and prescriptive analytics, it supports the digitalization of supply chain planning processes, automating decision making or replicating human decision-making processes to augment talent.” The “deep learning” to which Gartner analysts refer involves semantic reasoning and machine learning rather than neural networks.
Enterra has full confidence from this initial and important project work that the marriage of the ECS and SAP IBP can be used to demonstrate how to achieve all of these benefits in the years ahead. Although we will never be able to predict the future,” writes Eshkenazi, “with a bit of extrapolation and foresight, we can try to imagine it. Today, advances in science, technology and supply chain management may only be constrained by the limits of our imagination and ingenuity.” At the COIL, SAP and Enterra Solution are trying to stretch the limits of imagination. The more we experiment and co-innovate the more stretching we do. That’s why we are confident our efforts can contribute to forming a truly autonomous supply chain.